Mearls spent a decent amount of time putting together a scope for what he wanted in this Warlord design. He also did some number crunching to help ensure the design was balanced around something that already existed. These are good design practices. However, after using those two methods he begins to falter in his design process. The forming or art of creating the subclass is good, but once the form is roughed out, in my opinion, a designer should compare the form to the customer requirements or goals. He seems to fail in going back to his goals to ensure he is hitting them. In addition, a follow up step is to look at possible failure modes of this design. There are several possible modes of failure listed in this thread. If he takes the design and lists the modes of failure indicated here plus any additional concerns or modes of failure he or his team can think of, then lists how often they occur, whats the severity of their occurrence, and how easy it is to make the failure happen; he will be one step closer to having a design that I would consider ready for play tests. As it is, the design would be a first draft of a new mouse trap. He hasn't sat back and analyzed how the mouse trap may or may not work in the game environment, rules systems, with other characters, and with monsters. I also think he needs to take a look at the warlord using a P-diagram to determine: the inputs by the player and DM, the expected outputs by the player and DM, noise factors that influence performance, control factors that player has to help achieve the expected outputs, and what are some un-intended outputs or consequences of use with the design.
These design tools (failure mode and effects analysis and P-diagram) are cumbersome tools to use and not applicable to every design. I feel the warlord has so much weight associated with it after reading a lot of forum threads that a good design would require the use of them. As a comparison, I feel he has the acrobat design really close and would only need to improve the design by looking for failure modes by himself in the way that a typical designer would.